HAL Id: hal-00299464
https://hal.archives-ouvertes.fr/hal-00299464
Submitted on 12 Nov 2007
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of
sci-entific research documents, whether they are
pub-lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Very Weak Signals (VWS) detected by stacking method
according to different astronomical periodicities
(HiCum)
M. van Ruymbeke, P. Zhu, N. Cadicheanu, S. Naslin
To cite this version:
M. van Ruymbeke, P. Zhu, N. Cadicheanu, S. Naslin. Very Weak Signals (VWS) detected by stacking
method according to different astronomical periodicities (HiCum). Natural Hazards and Earth System
Science, Copernicus Publications on behalf of the European Geosciences Union, 2007, 7 (6),
pp.651-656. �hal-00299464�
www.nat-hazards-earth-syst-sci.net/7/651/2007/ © Author(s) 2007. This work is licensed under a Creative Commons License.
and Earth
System Sciences
Very Weak Signals (VWS) detected by stacking method according to
different astronomical periodicities (HiCum)
M. van Ruymbeke1, P. Zhu1, N. Cadicheanu2, and S. Naslin1
1Royal Observatory of Belgium, Belgium
2Institute of Geodynamics of Romanian Academy, Bucharest, Romania
Received: 20 July 2007 – Revised: 18 October 2007 – Accepted: 22 October 2007 – Published: 12 November 2007
Abstract. A stacking method to detect very weak signals is
introduced in this paper. This method is to stack observed data in different well known periodicities according to the astronomical clock since majority geophysical observations are time based. We validated this method by applying it in four different cases. Interactions behind the observed param-eters become obviously after it is stacked in two diurnal and semidiurnal tidal periodical waves. Amplitude and phase variations will be also measurable when a sliding windows stacking is used. This could be an important reference to find precursors before some earthquakes and volcanic events, cor-responding to attenuations of medium patterns.
1 Introduction
The new development of HiCum stacking method (His-togram Cumulating) is introduced in this paper (van Ruym-beke et al., 2003). Detect different periodical signals is an important issue in nature science, especially in the geophysi-cal domain. Periodicities of signals have been widely studied since Fourier transform algorithm was published. However, in our study, we are more focused on the properties of differ-ent signals inside the pre-defined periodicities. Earth is con-tinuously moving inside the gravitational gradient induced by the Moon and Sun. The orbital parameters of the motion are strictly defined by the celestial mechanics so that the reaction of the Earth to each outside body produces many periodical signals and most of them could be separated using stacking method if long enough time series are available. For instance, the lunar and solar semidiurnal wave could be separated with one month data. But, the ultimate purpose of the method is not only to separate the period signal of observed parameters, but also to detect the signal’s amplitude and phase variations
Correspondence to: P. Zhu
(zhooping@gmail.be)
in time domain and to connect them with past events such as earthquakes and volcanic events. For example, small vari-ations of amplitude distribution before a strong earthquake have been found when it is used to investigate the relation between tidal stress and Vrancea seimic zone (Cadicheanu et al., 20071).
Multi-parameters approach is required for the monitor-ing of complex interactions induced by tectonic activities in seismic, volcanic, geochemical areas, hydrology, land-slide zones. To illustrate local inter-activity between parameters, we analyzed four situations with HiCum, The barometric pressure induced on tidal gravity field variations in Membach (Belgium), sensor recording sea level in a lava tube in Lan-zarote (Spain), monitoring of strain, pressure and tempera-ture in underground cave at Rochefort (Belgium), and ground water level inside boreholes plunging in three aquifers at Uc-cle (Belgium). All of the examples show how the method works on each geophysical experiment.
2 Description of “HiCum” stacking method
Stacking data has been used for many decades in the meteo-rological fields (Bartels, 1938). In fact, as early as 19th cen-tury Darwin firstly used similar method to observe the sea tides cycle. Following that idea, Doodson further promoted the work and built the precisely sea tides forecast “Doodson-L´eg´e” machine, which was in daily use since 1948 until it was superseded by the electronic computer during 1960s. The stacking method introduced here is based on similar principle; it separates observed signals according to prede-fined periodicities. Time based observation is a combination of periodical signals with nonlinear factors, so that main fre-quencies of the contained signal could be localized by the
1Cadicheanu, N., van Ruymbeke, M., and Zhu, P.: Tidal
Trig-gering evidence of intermediate depth earthquakes in the Vrancea zone (Romania), Nat. Hazards Earth Syst. Sci., in review, 2007.
652 M. Ruymbeke et al.: Stacking detector on geophysical signals 0 100 200 300 −2000 0 2000 Obs(nm/s 2) 0 100 200 300 −2000 0 2000 Syn(nm/s 2) 0 100 200 300 −50 0 50 P(hPa) 0 100 200 300 −200 0 200 Res(nm/s 2) t(day) 1 2 3 4 5 100 105 1010 1 2 3 4 5 100 105 1010 1 2 3 4 5 100 105 1010 1 2 3 4 5 100 105 1010 Frequency(cpd)
Fig. 1. (a) 365 days SG data Membach (Belgium), Synthetic tides,
Gravity residuals, Barometric Pressure and (b) their corresponding Fourier spectrum.
Fourier Transform, an important hypothesis for the HiCum method being that the period is precisely priori defined. A periodical signal could be expressed as:
f (t ) = N s1 N s X i=1 T X j =1 y(tj) + ε i = 1, T , 2T , ...Ns j = 1, 2, 3, ....T (1) T =2π ω (2)
where f (t ) represents averaging stacking results, t the time, y(t ) the observed data, N s the stacking times, T the period, ω: the angular speed and ε the uncertainties and er-rors. The precision of the stacking results is closely related with the data length n, the stacking period T , and the sam-pling interval dt. Then stacking results can be fitted by a sinusoidal function.
s(t ) = A sin(ωtj+ α) (3)
where A is the amplitude and α the phase
A = v u u t T X j =1 ((f (tj) sin(ωtj))2+ (f (tj) cos(ωtj))2) (4) α = tan−1 T P j =1 f (tj) sin(ωtj) T P j =1 f (tj) cos(ωtj) (5) 0 T −400 −2000 200 400 0 T −400 −2000 200 400 0 T −0.05 0 0.05 0 T −0.2 0 0.2 0.4 0 T −200 0 200 0 T −200 0 200 0 T −0.04 −0.020 0.02 0.04 0 T −0.4 −0.20 0.2 0.4 0 T −200 0 200 0 T −200 0 200 0 T −0.02 0 0.02 0 T −0.2 0 0.2 0.4 0 T −100 0 100 0 T −100 0 100 0 T −0.2 0 0.2 0 T −1 −0.5 0 0.5
Fig. 2. HiCum Staking results of K1, O1, M2, S2 periods applied to
the four channels which are observed tides, synthetic tides, gravity residuals and barometric pressure.
The goodness of the fit could be valued by root mean square error RMS RMS= v u u t 1 T T X j =1 (f (tj)−s(tj))2 (6)
For instance, if one year gravity records with minute sampling is stacked in the lunar period M2 (central fre-quency 1.9504 cycle per day, corresponding to 745 min the period), the M2 wave could be separated by 705 (365.15 day*1440 min/745 min) times stacking and the mean square error will be reduced about√N s times (this example
is√705=26 times). Sometimes, for a high quality data bank with high sampling rate compare to the signal’s period, the signal could be easily separated by several stacks. It provides a possibility to check the amplitude and phase variation for the signal in an approximated time interval. It is an important approach for the stacking method because, in the tidal trig-gering evidence of earthquakes and volcanic events studies, the amplitude and phase distribution changes in the tidal pe-riod could be a possible precursor for the forecasting of these events (Tanaka et al., 2006; Nicoleta et al., 2007). So that, the static stacking function is developed into a dynamic one by a sliding windowed stacking function:
fw(t ) = N X k=1 f (t ) (7) fw(t )= N X k=1 " 1 N s N s X i=1 T X j =1 y(tj) + ε # k=1, w, 2w, 3w, ...N (8)
where fw(t ) is a windowed stacking result, N the
num-ber of the windows, w the sliding window length and w≥T . Nat. Hazards Earth Syst. Sci., 7, 651–656, 2007 www.nat-hazards-earth-syst-sci.net/7/651/2007/
Table 1. Selected astronomical periodicities and corresponding gravitational tides amplitudes for Belgium.
Symbol Period Amplitude Origin T(Min) (nm/s2)
K1 1436 431.2255 Lunar & Solar declinational wave O1 1549 306.6215 Lunar principal wave M2 745 314.8163 Lunar principal wave S2 720 146.4691 Solar principal wave
Through Eqs. (1), (3), (6), we can build up a matrix E with the stacking times N s and its corresponding stacking results errors RMS E = N s RMS (9)
The minimum stacking times N s could be determined by optimizing the RMS value. So a long series record could be divided into N continuous sliding windowed stacking, once the stacking time N s fixed. Sliding stacking results give the amplitude and phase variation at approximated time interval
w.
3 Membach superconducting gravimeter record (SGs)
We firstly selected a one year record of the Superconducting Gravimeter installed in the Membach station (Belgium). This station (50.6092◦N, 6.0067◦E, altitude 250 m) is among the
best Belgian seismic stations and is usual as the reference point for the Belgian gravity network. It consists of a 130 m long gallery excavated in low-porosity argillaceous sand-stone with quartzite beds.
We compared separately results of Fourier spectrum and HiCum stacking on 365 days data of theoretical tides, grav-ity residuals and barometric pressure (Fig. 1a). The synthetic tides was calculated by Tsoft with a local mode adjusted by long term SGs observations (Van Camp and Vauterin, 2005). The gravity residuals are directly computed by subtraction of the theoretical value from the observed tide. Diurnal, semi-diurnal and ter-semi-diurnal energies appear obviously in the spec-trum of observed and synthetic tides. There are very small peaks in the semi-diurnal and ter-diurnal frequency band of the residuals and barometric Fourier spectrum (Fig. 1b.) by which it’s difficult to figure out the barometric pressure ef-fect on the SGs record. However, the pressure efef-fect be-comes very clear in the S2 period (Table 1) stacking results when the same data set is stacked in by HiCum. We selected the four waves which share 70% total energy of gravity tides coming from lunar and solar attraction (Table 1). K1 is the strongest wave and the barometric and ocean loadings have the minimum effect on O1 wave nearby Membach station (Melchior, 1978). Two main semidiurnal waves M2 and S2 are analyzed too (Fig. 2). The linear regression between the
M2 Window(n) T(minute) 2 4 6 8 10 12 100 200 300 400 500 600 700 K1 Window(n) 2 4 6 8 10 12 200 400 600 800 1000 1200 1400
Fig. 3. Sliding windows stacking on K1 and M2 period, the beat
effect from gravitational force appeared in K1, and lunar node effect could be observed in M2.
HiCum stacking results on four periods, confirms the qual-ity of the synthetic mode adjustment with differences to the ideal unity (1 nm/s2) factor smaller than 0.1% except for S2
which could be more perturbed by the main tidal compo-nent existing in the barometric spectrum. We compare the atmospheric pressure record and HiCum stacking residuals. Correlation appears clearly for the K1 and S2 component. There is very small barometric effect on M2 but it is well correlated with the residuals, the slope of best fit S2 wave is 2.829950 nm/s2/hPa which is very near the global mode (3.0 nm/s2/hPa) for correction barometric pressure on SGs record (Table 2).
The sliding approach is applied on the K1 and M2 pe-riod. One year SGs records are stacked in a 30 days window and the window slided with 30 days step. For each window, we applied sliding stacking for the K1 and M2 periodicities (Fig. 3). The sequenced outputs show the change of the two periodicity waves in time domain. For M2 which is the main lunar principle wave, its amplitude should be nearly constant (Melchior, 1978). The 30 days windows allow separating the M2 component of the solar S2 one. The lunar orbit node is about 27.32 days which appears in the sliding stacking as a phase shift. The K1 period is totally different because it contains the action of solar and lunar declinations, so that appears the beat effect inside the K1 periodicity (Fig. 3).
654 M. Ruymbeke et al.: Stacking detector on geophysical signals
Table 2. Linear regressions between observed and synthetic tides, between residuals and barometric pressure.
Linear regression between observed Linear regression between residuals tides and synthetic tides (y=ax+b) and barometic pressure (y=ax+b)
a b R2 rms a b R2 rms K1 1.0003 −0.000628 0.999998 0.467 −2.756 −0.000866 0.895025 0.155 O1 0.9991 0.009490 1.000000 0.019 −3.215 0.006789 0.116549 0.224 M2 0.9992 0.003947 1.000000 0.131 −6.633 −0.004588 0.930038 0.063 S2 1.0017 0.000701 0.999971 0.668 −2.830 0.000100 0.987746 0.077 0 200 400 600 800 1000 1200 1400 1600 1800 −100 0 100 200 t(day) Obs(cm) 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 100 105 1010 Frequency(cpd) Obs(cm)
Fig. 4. (a), Jameos del Agua Sea-level sensor record on 5 years
in-terval and (b) its Fourier Spectrum which confirms the non-linearity pattern with existing peaks between the fundamental diurnal periods until the fifth harmonic.
We can conclude from this example: the goodness of the synthetic tides mode and the barometric effect on gravimeter could be found after staking in four periods and the lunar orbit node effect are obviously in the sliding stack results.
4 Lanzarote sea gauge
The next example is an EDAS sea-level monitoring station ,set-up at Lanzarote Island (Spain) which is the most north-eastern island of the Canarian Archipelago (van Ruymbeke, 2001) . The landscape of Lanzarote is dominated by numer-ous volcanoes. The site of observation named “Jameos del Agua” is located in a lava tunnel of the quaternary volcano La Corona (609 m). The geometry of this tunnel consists in a series of small lakes interconnected by passages for water. This tunnel is plunging into the sea for more than one and half kilometers. Oceanic tidal there could reach few meters. The level depends continuously of the water flow in the tun-nel which is function of the passage section. This is a typical situation of non linearity with transfer function influenced by the signal itself.
A sea gauge sensor records the sea level in the first open lake of the lava tunnel. Very smooth motion appears due to the filtering of short period sea waves through the few open-ings existing between the open-sea and the tunnel. Fourier spectrum confirms the nonlinearity of the water motion in-side the lava tunnel (Fig. 4b).
The HiCum is applied to five years records. M2 amplitude is 40.528 cm of sinusoidal fit with RMS error 0.8261 and S2 amplitude 4.972 cm of sinusoidal fit with RMS error 0.2132. It shows the efficiency of stacking to separate harmonic com-ponents on long registrations (Fig. 5).
5 Rochefort strain meter
This example concerns the detection of very weak signals (VWS) in a nosing signal of a strain meter. The instrument is located in the Rochefort karstic cave built-up by the erosion of the Lomme River (Belgium). The monitoring is assumed with several EDAS sensors. We apply the HiCum analysis on a file covering 800 day’s records with four channels. The first channel is a glass stain meter RA1 located on a vertical crack. The second and the third channels are two thermome-ters monitoring temperature of the surface rock and environ-ment in the gallery nearby the strain-meter. The fourth chan-nel is a barometer (Fig. 6a) (van Ruymbeke et al., 2004).
The spectral analysis of the four channels shows differ-ent peaks in diurnal, semi-diurnal and ter-diurnal frequency bands (Fig. 6b).
For the strainmeter, diurnal and semi-diurnal activities ap-pear clearly without harmonics. Contrarily for rock temper-ature and barometric pressure, semi-diurnal and ter-diurnal peaks exist and seem to be correlated. We subtract the raw dada with a third degree polynomial function to eliminate long term instrument drift before stacking the data (Fig. 7).
The length of strain meter is only 15 cm. We observe that the HiCum for the strainmeter on M2 and S2 are very well determined. The amplitudes reach about 20 nanometers with a noise level of only a few nanometers! On the 4th channel, HiCum confirms a significant S2 barometric component with a negligible M2 one. The air thermometer HiCum on S2 is like the derivative of the barometric modulation, confirming
0 T −30 −20 −10 0 10 20 30 40 A(cm) 0 T −4 −3 −2 −1 0 1 2 3 4 5
Fig. 5. HiCum of Jameos J31 sea gauge. HiCum results on M2 and
S2 periods. 0 200 400 600 800 −5 0 5 Obs( µ m) 1 2 3 4 5 100 105 0 200 400 600 800 −0.02 0 0.02 T( ° ) 1 2 3 4 5 10−5 100 105 0 200 400 600 800 −0.02 0 0.02 T( ° ) 1 2 3 4 5 10−5 100 105 0 200 400 600 800 1000 1020 1040 t(day) P(hPa) 1 2 3 4 5 100 105 1010 Frequency(cpd)
Fig. 6. (a) Four channels recording signals from strain-meter, air
thermometer, rock thermometer and barometer. (b) Corresponding Fourier spectrum.
the induction process found with spectral analysis (Rasson and van Ruymbeke, 1977).
No significant effect appears for M2. For rock thermome-ter, only residual effects exist which could be related to evap-oration of water flowing through cracks which could influ-ence the strain-meter record.
These graphs illustrate the interests of the HiCum stack-ing method for problems with complex interactions. The records of strain-meter are very noisy with a lot of irregu-larities. However, HiCum on long time interval permits to extract clear information with sufficiently large dynamics to validate physics models.
0 T −0.01 0 0.01 A( µ m) 0 T −2 0 2x 10 −5 T( ° ) 0 T −10−5 0 5 x 10−6 T( ° ) 0 T −5 0 5 x 10−3 P(hPa) 0 T −0.01 0 0.01 0 T −2 0 2x 10 −5 0 T −5 0 5x 10 −5 0 T −0.05 0 0.05
Fig. 7. HiCum applied to the raw data of four channels
(strain-meter, rock thermo(strain-meter, air thermometer and barometer) for lunar M2 period and solar S2 period.
0 200 400 600 −10 0 10 B1( µ m) 1 2 3 4 5 100 105 0 200 400 600 −10 0 10 B2( µ m) 1 2 3 4 5 100 105 1010 0 200 400 600 −50 0 50 B3( µ m) 1 2 3 4 5 100 105 1010 0 200 400 600 950 1000 1050 t(day) P(hPa) 1 2 3 4 5 100 105 1010 Frequency(cpd)
Fig. 8. (a) The four channels show the signals of the three boreholes
water levels and, barometric pressure. (b) the corresponding Fourier spectrum.
6 Uccl boreholes water levels
We illustrate a medium interaction of actions by analyzing a borehole ground water level station located at the Royal Observatory of Belgium. It records the change of the water levels in three tubes plunging separately in three independent aquifers located at different depths (B1: −35 m, B2: −64 m, B3: −95 m). The upper aquifer is an opened one, the sec-ond a semi-confined, and the third a confined one (Fig. 8a) (van Ruymbeke and Delcourt, 1985). The original records were pre-treated with the same procedure as the previously studied Rochefort one. Main conclusions from the spectral analysis concern interaction of water levels with simultane-ous atmospheric pressure and tidal gravity changes. However the signal to noise ratio for spectrums is too low for under-standing the relations between these two effects (Fig. 8b).
656 M. Ruymbeke et al.: Stacking detector on geophysical signals 0 T −2 0 2 x 10−3 B1(cm) 0 T −5 0 5 x 10−3 B2(cm) 0 T −0.05 0 0.05 B3(cm) 0 T −0.02 0 0.02 P(hPa) 0 T −0.05 0 0.05 0 T −5 0 5 x 10−3 0 T −0.05 0 0.05 0 T −0.2 0 0.2
Fig. 9. HiCum with lunar M2 and solar S2 period graphs applied to
the four channels (water levels B1, B2 & B3 and barometer).
We apply the HiCum method on same series of data. For the first borehole and the barometric channel, HiCum shows a common large amplitude S2 and negligible modulations on M2. The gravitational effect appears on the B3-aquifer and smoothly on B1. Induction of atmospheric pressure looks negligible for the B2 and B3 boreholes. It is another example of this method to compare by stacking, two kinds of actions by the influences detected in reactions (Fig. 9).
7 Discussion
This paper has introduced a stacking method (HiCum) which could transform time based observation into different period-ical based signals by stacking. One important issue in Earth science is to find periodical phenomena and to explain the physical meanings behind it. HiCum stacking method is an-other way to precisely study the performance of different sig-nals in the frame of a known period T . The synthetic tide mode accuracy and barometric effect on gravimeter could be determined when the signals are stacked in four periods. Sliding stacking shows the well knows lunar node effect in the form of phase shift inside each window. In the sea gauge observations, the main semidiurnal lunar and solar tides sig-nals could be subtracted and separated; for the strain meter records, very small tidal inductions can be detected even if the record is largely polluted by moisture, pressure and tem-perature. In the borehole water level monitoring stations, the semidiurnal tides signals are detected in the two deep-est aquifers. Above all, these examples are just illustrating how the stacking method works. However, several aspects of the method still need to be investigated specially: its perfor-mance on short period signal studies and its adaptabilities in other domains.
Acknowledgements. The authors are grateful to all people who
participate in EDAS projects; we thank particularly M. van Camp and M. Hendrickx for giving the Membach SG data, and E. de Kerchove correcting manuscript. The experiments in Lanzarote were organized with the support of R. Vieira and his colleagues; Local facilities are purchased by the Casa de los Volcanoes belongs to the Cabildo Insular de Lanzarote. G. Tuts has prepared the data files. The project in Rochefort was supported by Y. Quinif and Th. Camelbeeck. This research is financially supported in part by the Action 2 contract from the Belgian Ministry of Scientific Politics. We deeply appreciate the comments from B. Meurers and anonymous reviewer that served improves significantly the original manuscript.
Edited by: P. F. Biagi
Reviewed by: B. Meurers and another anonymous referee
References
Bartels, J. : Random Fluctuations, Persistence and Quasi-persistence in Geophysical and Cosmical Periodicities , Terr. Magn. Atmos. Electricity, 40, 1, 1–60, 1938.
Melchior, P.: The Tides of the Planet Earth, Pergamon Press, Lon-don, 1978.
Rasson, J. and van Ruymbeke, M.: Underground meteorology at the underground Laboratory of Geodynamics at Walverdange, in: VIIIth Symp. on Earth Tides, Bonn, 1977.
Tanaka, S., Sato, H., Matsumura, S., and Ohtake, M.: Tidal trigger-ing of earthquakes in the subducttrigger-ing Philippine Sea plate beneath the locked zone of the plate interface in the Tokai region, Japan, Tectonophysics, 417, 69–80, 2006.
van Camp, M. and Vauterin, P.: Tsoft: graphical and interactive software for the analysis of time series and Earth tides, Comput. Geosci., 31(5) 631–640, 2005.
van Ruymbeke, M., Howard, R., P¨utz, P., Beauducel, F., Somer-hausen, A. and Barriot, J-P.: An Introduction to the use of HiCum for signal analysis, BIM 138, 10 955–10 966, 2003.
van Ruymbeke, M.: Projects of the Royal Observatory of Bel-gium (ROB) at the Lanzarote Geodynamical Laboratory (LGL), J. Geodetic Society of Japan, 47, 463–469, 2001.
van Ruymbeke, M. : The Environmental Data Acquisition Sys-tem(EDAS) developed at the Royal Observatory of Belgium, J. Geodetic Society of Japan, 47, 40–46, 2001.
van Ruymbeke, M. and M. Delcourt : A capacitive transducer for water level,International Symposium on Earth Tides, Consejo Superior de Investigaciones Cientificas, Madrid, 95–103, 1985. van Ruymbeke, M., Beauducel, M. Fr., and Somerhausen, A.: The
Environmental Data Acquisition System(EDAS). Developed at the Royal Observatory of Belgium, in: Short thermal and hy-drological signatures related to tectonic activity, Cahier du Cen-tre European de Geodynamique et de Seimologie, 14, 163–174, 1997.
van Ruymbeke, M., Shaoming, L., Quinif, Y., and Camelbeeck, T.: The Monitoring of Tectonic Movements In Nature Caves, the 3rd International Conference on Continental Earthquakes Mech-anism, Prediction, Emergency Management and Insurance, Bei-jing, 134–140, 2004.
